Research On Structure System Reliability Based On Iterative Adaptive Response Surface Method

Among the traditional response surface methods (RSM), the quadratic RSMwithout cross terms and the totally quadratic polynomial RSM are most widely used.However, both efficiency and accuracy can’t all be considered at the same time. Theadaptive RSM considering cross terms can achieve an equilibrium point between theefficiency and accuracy. In order to improve the accuracy deeply, the adaptive RSMconsidering cross terms is improved by introducing the idea of updating sample centeriteratively and convergence criteria. Four different numerical examples are analyzed,the results performing that the adaptive RSM with cross terms can achieve highefficiency relatively, and the proposed improved method can improve the accuracy ofcalculation results.Most of the structure reliability analysis methods based on the RSM are aimed atcomponent reliability of single failure mode, which contains only one function and canget relatively ideal results. However, the classical structure system reliability and thegeneral structure system reliability often have multiple performance functions. Multipleresponse surfaces method is often processed by scholars, thus the advantages of RSM inefficiency and accuracy are lost. Based on the theory of equivalent extreme value event,this paper transforms multiple functions of structure system into single function, whichmakes the RSM can be well used in the structure system reliability analysis efficiently.Therefore, based on the combination between the improved iterative adaptive RSM andthe theory of equivalent extreme value event, a method for the structure systemreliability analysis is proposed, for which implementation steps are designed. In thissection, five different examples are analyzed to verify the proposed algorithm, and theresults show that it is accurate and effective.For a practical grid structure, the finite element mold is builded firstly. The resultsof structure static analysis show that the structure has good performance in linearanalysis and high bearing capacity in nonlinear state. On the other side, four differentkinds of structure system reliability are analyzed by using the proposed algorithm in thispaper, and results show that the structure has a higher reliability. In this section, theinfluence of random variable snow load on the structure reliability is also analyzed, andresults show that average of snow load and snow distribution coefficient are influentialgreatly. Last, this paper makes a preliminary researching on the relationship between thecomponent reliability and the structure system reliability, which also can be seen as theexploratory research of the structural optimization method based on the structure systemreliability or structure design method directly based on the system reliability. Thealgorithm for system reliability analysis which be recommended in this paper, iscombined with structure optimization based on component reliability theory, in order tosduty the relationship between reliability index of the lowest component constraint limitwith the system reliability index of the optimized structures. The results show that therelationship between the two indexes is linear approximately, which is also given out inthis section.